Hero BackgroundHero PatternHero Pattern

Blog

Latest insights and updates

Back to Articles
AI Solutions

Automation & AI in Commerce (2025): From Chatbots to Agentic Operations

2025-09-018 minutes
AI Commerce Revolution 2025

Key takeaways

  • AI shopping assistants are going mainstream. Walmart says its assistant Sparky will become the primary vehicle for discovery, shopping, and managing tasks like reorders and returns.
  • Amazon's Rufus is available to all U.S. customers, embedding conversational help directly in the app and desktop.
  • In Europe, the EU AI Act began applying GPAI transparency rules on August 2, 2025; a Code of Practice for model providers launched in July to help with compliance.
  • Retail adoption is accelerating, especially in customer service, operations, and marketing, but most leaders still want humans-in-the-loop for complex decisions.

What's new in 2025

Shopping becomes conversational. Walmart's CEO expects Sparky to be the company's primary digital vehicle for shopping and post-purchase tasks; the assistant is being built as a true retail companion.
Rufus now guides discovery and comparisons inside Amazon's core experience.

Regulation gets real. In the EU, general-purpose AI (GPAI) providers must meet transparency and copyright obligations now, with additional requirements for systemic-risk models; the GPAI Code of Practice offers a practical path to show AI Act alignment.

Adoption (but with guardrails). UK surveys show retailers actively deploying AI most cite gains in support, ops, and marketing, while keeping humans-in-the-loop on critical steps. Exploration of autonomous agents is rising fast.

Where value is happening (today)

1. Customer service, reimagined. Agentic assistants understand requests, retrieve policy/order data, and act (returns, refunds, rebookings), handing off to agents when needed. KPIs: first response time (FRT)↓, average handle time (AHT)↓, customer satisfaction (CSAT)↑.

2. Search & discovery. In-journey assistants increase product findability, reduce pogo-sticking, and lift conversion/AOV, rufus is the reference point at scale.

3. Inventory & supply. Retailers are moving from reactive to predictive: granular forecasts, anomaly detection, and smarter replenishment to cut stockouts/overstock.

4. Associate assist. Store and contact-center employees use AI to answer faster, translate, and follow procedures, less swivel-chair work, more selling.

5. Marketing & retail media. Generative workflows create variants, match intent, and optimize budget, focus shifts from hype to ROI.

Quick playbook for SMEs (60–90 days)

Weeks 0–2: Pick one journey + three KPIs
Choose a high-volume path (FAQs/returns or product discovery) and commit to 3 metrics (e.g., FRT, conversion, cost-per-contact).

Weeks 3–6: Ship a connected pilot
Deploy a conversational agent wired to your catalog, orders, and help desk/CRM. Add guardrails (tone, policy blocks) and human handoff.

Weeks 7–10: Ground in your data
Load policies/FAQs, product docs, and workflows; enable event logging for audits and consent prompts on data use.

Weeks 11–13: Test, govern, extend
A/B test vs. current flows. Align with AI Act basics, transparency, risk management, human oversight. Expand to WhatsApp/web once KPIs move.

Governance & compliance (EU AI Act—Fast checklist)

  • Say it clearly: disclose when users interact with AI; keep humans-in-the-loop for impactful decisions.
  • Manage risk: document use-cases, evaluate failure modes, monitor drift; keep an incident process.
  • Respect rights: apply data minimization and retention; be ready for access/deletion requests.
  • For GPAI vendors/models you use: prefer those aligned to the GPAI Code of Practice.

What to measure

What to measure in AI Commerce
  • Experience: conversion, AOV, and CSAT/NPS.
  • Support efficiency: FRT, AHT, deflection/self-service rate, and cost per contact.
  • Operations: stockouts/overstock, forecast error (MAPE), and time-to-resolution.
  • Risk: escalations to human, policy-violation rate, and incident count.

Conclusion

2025 is the year automation in commerce turns fully practical: assistants in the shopping journey, agentic ops in service and supply, and compliance frameworks that clarify the rules. The winners pick one journey, wire AI to real systems and data, measure relentlessly, then scale.

Next step

Want a 15-minute walkthrough to scope a pilot for your store or marketplace?
Email: contact@tbeninnovation.com

Share it!

Continue Reading

How AI Solves Key Challenges in Customer Service and Commerce

Read Next Article